A New Method of Feature Selection for Information Fusion Based on the Game Theory

This paper extracts the player and strategy set from feature space by using the game theory. In doing so, it depicts the failure to define feature information by using mutual information entropy. It also exposes the underlying nature of the decision-making process that is founded on conflict and cooperation through the construction of a payment function. Finally, the paper seeks to provide a balance solution for this strategy by the payoff matrix so that the optimum feature subset can be obtained. This feature subset can improve the performance of the decision-making system; hence, the feature dimension catastrophe caused by the high feature dimension can be subsequently addressed. Consequently, after the above-mentioned methods have been used to conduct feature selection as applied in vehicle selection, it may stated that vehicle feature dimensionality can be compressed. The recognition efficiency can also be improved greatly.